Report Local governance, decentralisation and corruption in Bangladesh and Nigeria

Transcription

1 Report Local governance, decentralisation and corruption in Bangladesh and Nigeria Hamish A. D. Nixon and Alina Rocha Menocal, Nieves Zúñiga, Debapriya Bhattacharya, Syed Muhtasim Fuad, Idayat Hassan, Kelechi C. Iwuamadi, Umme Shefa Rezbana, and Shamsudeen Yusuf April 2018

2 This material has been funded by the British Academy through UK aid from the UK government. However, the views expressed do not necessarily reflect the British Academy's or the UK government s official policies. Overseas Development Institute 203 Blackfriars Road London SE1 8NJ Tel: +44 (0) Fax: +44 (0) info@odi.org.uk Readers are encouraged to reproduce material from ODI publications for their own outputs, as long as they are not being sold commercially. As copyright holder, ODI requests due acknowledgement and a copy of the publication. For online use, we ask readers to link to the original resource on the ODI website. The views presented in this paper are those of the author(s) and do not necessarily represent the views of ODI. Overseas Development Institute This work is licensed under a Creative Commons Attribution-NonCommercial Licence (CC BY-NC 4.0). Cover photo: Selling freshly caught fish in Orimedu, Lagos State, Nigeria Arne Hoel/World Bank, 2010

3 About this report This report synthesises the findings and implications of a two-year research collaboration between the Overseas Development Institute (ODI) in the UK, the Centre for Policy Dialogue (CPD) in Bangladesh, and the Centre for Democracy and Development (CDD) in Nigeria on Multi-level Governance, Decentralisation and Corruption. The study formed part of the Anti-Corruption Evidence Grant Programme (ACE) funded by the UK Department for International Development (DFID) and administered by the British Academy. Acknowledgements The ODI study team, Centre for Policy Dialogue, and Centre for Democracy and Development wish to acknowledge the financial support received under the UK Department for International Development funded Anti-Corruption Evidence Programme (ACE), administered by the British Academy. The team is deeply grateful to all the people they interviewed and spoke with as part of this project for their time and insights. The study has benefited from collaboration and interaction between the ACE study teams facilitated by the British Academy, and the leadership of Professor Paul Heywood in guiding the ACE grants program. The team thanks fellow panel members and audiences at the Development Studies Association annual conference in Oxford in 2016 (Caryn Pfeiffer of the University of Bristol, Dominik Zaum of the University of Reading, Rosita Armytage of the Australian National University and Alice Evans of King s College, Cambridge) and the Australasian Aid Conference in Canberra in 2018 (Professor Heather Marquette from the University of Birmingham, Lili Mark from the Central European University, Kristian Futol of the Department of Foreign Affairs and Trade and Professor Heywood) for their comments and contributions. The team also wish to thank Professor Paul Smoke of New York University for comments at various points in the design of the research, and Professor Michael Johnston from Colgate University for his review and invaluable comment on this draft. Finally, a big thank you to Erin Shepherd, Hannah Caddick, Caelin Robinon, Helen Dempster, and George Richards at ODI for their essential support to this project throughout. All errors and omissions remain the authors alone. 3

4 Contents About this report 3 Acknowledgements 3 List of boxes, tables and figures 6 Acronyms 7 1 Introduction Research objectives Conceptual framework 9 2 Approach and methods Project approach Case selection Sector focus Data 13 3 Understanding decentralisation and corruption Empirical evidence on decentralisation, corruption and anti-corruption Approaches to understanding decentralisation Approaches to understanding corruption Implications for the relationships between decentralisation and corruption Implications for anti-corruption programming 24 4 Findings on decentralisation and corruption Corruption in a decentralised governance context Impacts of local corruption Causes of decentralised corruption Social norms around corruption Linking theory to reality: explaining local corruption 32 5 Findings on fighting corruption in decentralised governance Decentralised anti-corruption initiatives Fighting local corruption 35 4

5 6 Conclusions and implications What have we learnt about decentralised governance and local corruption? Improving the effectiveness of anti-corruption initiatives at decentralised levels A note on further research 38 References 39 Annex 1 Detailed research questions 42 5

6 List of boxes, tables and figures Boxes Box 1 The British Academy Anti-Corruption Evidence Programme 8 Box 2 Anti-corruption initiatives analysed as part of this project 10 Box 3 Data sources 15 Box 4 Dimensions and forms of decentralisation 17 Box 5 Four pillars of fiscal decentralisation 18 Box 6 Principal agent problems 18 Tables Table 1 Nigeria within-country case selection logic (12 LGAs) 12 Table 2 Bangladesh within-country case selection logic 13 Table 3 Tree approaches to decentralisation 17 Table 4 Causes, forms and manifestations of different decentralisation processes 19 Table 5 Different forms of corruption 20 Table 6 Theoretical approaches to relationships between decentralisation and corruption 23 Table 7 Reorienting theoretical perspectives linking decentralisation and corruption 36 Figures Figure 1 Research Framework 9 Figure 2 Nigeria study sites 12 Figure 3 Bangladesh study sites 14 6

7 Acronyms ACC ACE BA CPCs DFID EFCC FAAC FCT ICPC INEC LDCs LGA LGI MPs PFM RLGIs SIEC SLIP SMBs/SMCs TLM ULGIs UNCAC UP UZGP UZP Anti-Corruption Commission DFID s Anti-Corruption Evidence British Academy Corruption Prevention Committees UK Department for International Development Economic and Financial Crimes Commission Federation Account Allocation Committee Federal Capital Territory Independent Corrupt Practices and Related Matters Commission Independent National Electoral Commission Local Development Centres Local Government Area Local Government Institutions Members of Parliament public financial management Rural Local Government Institutions State Independent Election Commissions School Learning Improvement Plan School Management Committees Teaching-Learning Materials Urban Local Government Institutions United Nations Convention Against Corruption Union Parishad Upazila Parishad Governance Project Upazila Parishad 7

8 1 Introduction Corruption is high on the agenda of national governments, international organisations, aid providers and civil society. At the same time, within the context of broader democratization trends, decentralisation has become a dominant policy reform across the developing world. This generated enormous enthusiasm both domestically and within the international community about prospects for improved governance. Democratic decentralisation appears to promise government and services that are closer to the people, while fostering greater citizen voice and participation and increased accountability. The expectation is, in turn, that this will also help to reduce corruption. But in fact, some of the most prevalent manifestations of corruption are at the local level, and it is not clear that more localised forms of government will necessarily be less corrupt. The links between decentralisation and corruption are complex, and the role of decentralised governance in combatting corruption remains uncertain. The role of decentralised governance in combatting corruption remains uncertain and Corruption. The research project is part of the UK Department for International Development (DFID) funded Anti-Corruption Evidence Grant Programme (ACE) administered by the British Academy (Box 1). The research is based on a thorough review of the literature on corruption and decentralisation as well as two country studies on Bangladesh and Nigeria which have been published separately by CPD and CDD. These country studies are available via the ODI study page at Research objectives This research aims to deepen understanding of the links between decentralised governance and corruption, and what this implies for the effectiveness of anti-corruption measures at the local level by exploring these connections in Bangladesh and Nigeria. To do so, it explores the performance of decentralised governance arrangements in particular settings, analyses the nature and dynamics of corruption in those settings, and identifies implications of both for the effectiveness of anti-corruption efforts at local levels. Despite the efforts of both researchers and policy-makers to better understand the role decentralised governance and decentralisation reforms play in either generating and sustaining corruption or helping to combat it, there is no theoretical consensus or consistent empirical evidence on how corruption and decentralised governance affect one another. Corruption research and policymaking increasingly recognises the need to disaggregate corruption corruption takes many different forms and has many different causes and effects in different settings, and strategies to combat corruption are also likely vary across these types and settings. In addition, as this study highlights, different theoretical perspectives on both decentralisation and corruption can influence the way that their relationship is viewed and measured. This report synthesises the findings and implications of a two-year research collaboration between the Overseas Development Institute (ODI) in London, the Centre for Policy Dialogue (CPD) in Dhaka, Bangladesh, and the Centre for Democracy and Development (CDD) in Abuja, Nigeria, on Multi-level Governance, Decentralisation Box 1 The British Academy Anti-Corruption Evidence Programme The British Academy (BA) funded eight projects as part of a 4 million global anti-corruption research scheme in partnership with the UK Department for International Development (DFID). The BA-DFID partnership is one component of DFID s Anti-Corruption Evidence (ACE) Research Programme. The scheme supported eight research teams, whose work focuses primarily on DFID priority countries where corruption is a major constraint (including Ghana, Uganda, Kenya, Nigeria, Tanzania and Bangladesh). These projects identify new initiatives that can help developing countries tackle corruption and the negative impact it has on millions of people s lives. For more information, please visit the BA/ DFID ACE Partnership website: 8

9 Within this framework, the research focused on the following overarching questions: What is the nature of the governance relationships in each decentralised context? How does the multi-level governance system perform and why? How does corruption manifest itself and vary across different contexts, and how does corruption affect the sectors and functions identified and why? What approaches are being taken to anti-corruption in Bangladesh and Nigeria, how effective are they at local levels, and why? The research findings suggest a strong two-way relationship between the ways in which decentralised governance functions and the forms of corruption that are observed in local settings. The findings also shed light on the impact of local anti-corruption initiatives, and suggest that, if such measures are to be more effective they need to address the weaknesses and incoherence in the system of decentralised governance as it is implemented in a given setting. Such a focus that has thus far been rare in anti-corruption efforts in both Bangladesh and Nigeria. These relationships are illustrated by the arrows in the figure below. This study makes three broad contributions to the existing evidence in the area of corruption and decentralisation. First, we argue that the theoretical underpinnings of most analysis of the linkages between decentralised governance and corruption are incentive- and principal/agent-based. As such, these analyses of local corruption and possibly the means to tackle it do not respond to the full range of drivers of that corruption, especially more systemic ones like the implementation of decentralised local governance. Secondly, we suggest that analysis of corruption in decentralised and local settings should be reoriented towards approaches that privilege: (a) the political economy of decentralised systems, including the coherence of policy and implementation, within a broader system of governance; and (b) collective and not simply individual determinants of corrupt behaviour, such as incentives among levels of government, and how these may or may not translate into functional or social norms. Finally, building on these findings, the study helps to explain weaknesses in local anti-corruption measures analysed as part of the fieldwork for this project. Our research suggests that anti-corruption efforts at the local level can be strengthened by supplementing existing national agency, legal and regulatory, and/or awareness-based approaches with greater attention to the coherence of decentralised governance. Thus anticorruption initiatives should aim to redress weaknesses in that coherence rather than try to introduce new laws, regulations or awareness-raising measures. The sections below elaborate the conceptual framework, approach and methods developed and used to undertake this research in greater detail. 1.2 Conceptual framework As described in Figure 1, we developed a conceptual framework to anchor this research project that focuses on three core concepts decentralised governance, corruption and anti-corruption and the linkages between them. The conceptual framework helped to draw out more detailed research questions for each of the three concepts, and the dynamics between them (see Annex 1). Figure 1 Research Framework Corruption How does corruption manifest itself and how does it vary across different contexts? How does corruption affect sectors and functions identified and why? Decentralised governance What is the nature of the governance relationshipsin a given decentralised context? How does the multi-level governance system perform and why? Anti-corruption What approaches are being taken to anti-corruptionin Nigeria and Bangladesh, how effective are they at local levels, and why? 9

10 1.2.1 Decentralised governance Following Paul Smoke (2015a), this study interprets decentralised governance as: the assignment of public functions to subnational governments along with structures, systems, resources, and processes and that support implementing these functions to meet specific public sector goals. As such, the study focuses on understanding the effects of different arrangements of public functions across levels of government. In this kind of multi-level governance approach: the key underlying question is not whether to decentralize or not or even opt for a specific decentralization model, but to look at ways to improve capacity and co-ordination among public stakeholders. (Charbit, 2011a: 3) This approach to decentralised governance also demonstrates the importance of understanding how formal and informal institutions interact by highlighting different aspects of the political economy of decentralisation, including policy coherence and implementation Corruption The project approaches corruption with two important orientations: The study is inductive in that it focuses on learning from qualitative fieldwork the different forms that corruption takes in decentralised contexts, and seeing if these align with existing typologies, rather than testing for the presence and causes of a predetermined form of corruption. It takes a disaggregated approach to corruption: rather than focusing on general levels of corruption, the research has focused on corruption within particular areas of public sector activity, as detailed in the methodology section. The literature on corruption recognises that broad, often perception-based, measures of corruption can obscure the variety of forms and dynamics present in a given context, and these are key to understanding how to combat corruption. Indeed, as the broader literature on corruption and anti-corruption suggests, the issue of whether generic definitions of corruption can make a significant contribution to efforts to both understand and tackle it remains an open question (Rocha Menocal et al. 2015; Khan et al. 2016; Heywood 2017). This study takes an open and non-normative approach to conceptualising corruption, so as to capture differing manifestations of, and social norms and views about, corruption. Rather than adopting a generic definition of corruption this study focuses on corruption in the public sector at the local level, and explores abuses that are for private, but also other collective or party-political gains. For the purposes of this project, we have adopted on a functional definition of corruption that is based on the abuse of public authority for non-public advantage. This captures the role of public authority and the distortionary effects of corruption while remaining open enough for the research to be guided by the perspectives of respondents on what constitutes corruption. It avoids the use of private gain found in other definitions so that perspectives on corrupt gains, such as political or communal, may be captured. This definition is not intended to supplant or contest the wider generic definitions found in the broader anti-corruption literature (see Johnson et al and Rocha Menocal et al among others) Anti-corruption This concept is important to connect the research to what works and why in fighting corruption. However, there is no clear definition of an anti-corruption initiative. Anti-corruption initiatives may include those explicitly framed and aimed at fighting corruption (direct approaches), reforms with broader focus but plausible anti-corruption impact (indirect approaches), and the anti-corruption effects of given institutional arrangements (structural approaches) (Johnson et al. 2012). The approach of this study was to try to explore each of these types of anti-corruption approaches, and where possible, even the interactions between them. An important question is the role of local governance arrangements themselves (structural arrangements) in driving or fighting corruption. This research set out to explore three anti-corruption initiatives in different local settings in each country of study (Box 2). The initiatives were selected on the basis of several criteria, including that they: (a) capture plausible local impact; and (b) provide a comparison of both direct and indirect approaches to combat corruption. Box 2 Anti-corruption initiatives analysed as part of this project Nigeria Local Government Civil Service Rule Auditor-General of Local Government School-based management committees Bangladesh Anti-Corruption Commission outreach at the local level Transparency International Bangladesh Conscious citizens committees School Management Committees 10

11 2 Approach and methods This research explores the concepts and relationships described in Chapter 1 through inductive and primarily qualitative field research in two countries, Bangladesh and Nigeria. 2.1 Project approach Within both Bangladesh and Nigeria we studied several local administrations to highlight variation in both institutional and contextual dimensions. We paid particular attention to basic education and local public financial management to allow more detailed examination of particular corruption dynamics and their effects on the population at the local level. Finally, we explored what anti-corruption approaches may be relevant at the local level, how effective they have proven in combating corruption, and why. The research was guided by an approach that is: Disaggregated. Both decentralisation and corruption research identifies insufficient disaggregation as a barrier to understanding how corruption and decentralisation relate to each other. This study focuses on different forms of decentralised governance arrangements and corruption, rather than adopting a universal approach. It also looks at anti-corruption initiatives through this disaggregated lens. Inductive. In line with the first principle, this study aims to better understand the concepts and relationships it studies by observing them on the ground, rather than testing one or a small number of explanatory hypotheses. Theoretically pluralist. This study maintains an open mind about which theoretical models or explanations may offer insight into corruption at decentralised levels, and addresses the main strands of theory in its analysis. Multi-level. This study considers interconnections between different levels, from the local to the national. For example, local government capacity cannot be understood without the functioning of intergovernmental relationships, and corruption has a variety of drivers that run across different levels. Collaborative. This study has prioritised ownership and participation in the design, conduct and analysis of research by all partners. This approach has at times made establishing a common conceptual approach, analyses or comparisons more difficult and time consuming, but it is indispensable to build capacity and longer-term impact in partner countries. 2.2 Case selection Bangladesh and Nigeria, which are both priority countries for DFID, embody two very different models of decentralised governance (see Chapter 3). As such, the case selection has been purposeful, and is not intended to be representative or selected on the dependent variable (e.g. high or low levels or certain types of corruption). Neither are we intending to compare Bangladesh and Nigeria in the strict sense of that term. Rather, the selection is intended to highlight key dimensions of variation in potentially important independent variables and to draw out lessons that may emerge across the case studies. For example, the case selection helps to highlight the distinction between a heavily devolved and a heavily deconcentrated approach to decentralised governance, but we do not use this as an isolated explanatory factor; similarly, our research explores rural urban distinctions within each country without seeking to use these as an isolated explanatory factor. We selected local units to study within countries on the basis of two broad types of factors that help to draw out contextual differences in governance structures between the two countries. One category of explanatory factor relates to the institutionalisation and structure of governance institutions and relationships. The second category captures more structural or background characteristics of each setting. Below we explain how we operationalised these categories of criteria Within-country case selection for Nigeria In Nigeria, the governance variable we identified to capture different forms of accountability and institutionalisation is whether local government elections have been held. In Nigeria there are Local Government Areas (LGAs) where elections have been held consecutively since the restoration of democracy in 1999 and areas where they have been inconsistently held and are now suspended. Secondly, in Nigeria we compare urban and rural LGAs to capture key structural factors of interest: population, density, diversity. We captured these two dimensions across and within states following Nigeria s federal structure (Table 1). This approach yielded the following locations within which to explore LGAs that capture the elected/non-elected variation: Lagos, Enugu, and Federal Capital Territory (FCT). In each of the three states, local governments were further selected based on rural urban differences. Within the timeframe of the fieldwork, Abuja municipal (urban), Gwagwalada and Abaji area councils were covered within the FCT and governed by elected officials; 11

12 Table 1 Nigeria within-country case selection logic (12 LGAs) Nigeria Lagos Rural Urban Enugu Rural Urban FCT Rural Urban Elections 2 LGAs 2 LGAs 2 LGAs 1 LGA No recent elections 3 LGAs 1 LGA Enugu East, Isi-Uzo, Awgu, Enugu North and Nsukka local government areas were covered in Enugu state and governed by appointed officials; while Ejigbo, Ikeja, Oshodi-Isolo, and Eti-Osa local government areas were covered in Lagos state and governed by elected officials. Lagos State is in South West Nigeria, and is the commercial capital of Nigeria. The state has 20 LGAs, and is divided into the metropolitan city which covers 16 local governments and the non-metropolitan city which has four. Lagos State has generally been very consistent in conducting local government elections and, during our fieldwork, elected local government officials were running the local councils examined. Data collection took place primarily in Ejigbo, Ikeja, Eti-Osa, and Oshodi-Isolo LGAs. Enugu State is in South East Nigeria. It is known as Coal City because of abundant coal deposits across the state. The state is irregular in the conduct of local government elections and, at the time of fieldwork, there were nonelected officials running the affairs of the local councils. The state has 17 LGAs, most of which are rural. The fieldwork took place in Enugu East, Isi-Uzo, Awgu, Nsukka and Enugu North LGAs, and within the state capital. Federal Capital Territory (FCT) is in the northern central part of Nigeria. It is also known as Abuja, the political capital of Nigeria. The Independent National Electoral Commission (INEC) and its FCT administration has been consistent in conducting elections for the area councils. So, at the period of fieldwork, there were elected officials in the six area councils that make up the FCT. Data collection was undertaken in three out of the six area councils: Abuja Municipal, Gwagwalada, and Abaji. Figure 2 Nigeria study sites Sokoto Kebbi Zamfara Katsina Kano Jigawa Yobe Borno Kaduna Bauchi Gombe Niger Adamawa Oyo Ogun Osun Kwara Ekiti Ondo FCT Kogi Nassarawa Benue Plateau Taraba Lagos Edo Enugu Anambra Ebonyi Delta Imo Abia CrossRiver Bayelsa Rivers Akwa Ibom 12

13 2.2.2 Within-country case selection for Bangladesh In Bangladesh, we developed different selection procedures, as there is no variance on the presence of elections. However, institutionalisation of local government is still of interest, and so here we selected local settings according to the proximity of the Local Government Institutions (LGI) to sub-district centres. These considerations aimed to capture differences in the level of institutionalisation of local administration, to explore whether this impacts corruption. In Bangladesh, we covered two rural local government tiers the local Union Parishad (UP) level and the sub-district Upazila Parishad (UZP) level as well as parallel urban municipality structures. As in Nigeria, the structural dimension is thus represented through a rural urban distinction, but it also explores differences between two rural administrative tiers. Research for the project was conducted in six areas, including four Rural Local Government Institutions (RLGIs) and two Urban Local Government Institutions (ULGIs). We used three criteria to select the RLGIs: distance from Sadar (the centre area at the local level); performance under the Upazila Parishad Governance Project (UZGP) (for the UZPs); and performance under the Union Parishad Governance Project (UPGP) (for the UPs). These two support projects active from 2011 to 2016 focused on strengthening the UZPs and UPs to be more functional, democratic, transparent and accountable institutions. Following these criteria, we selected four rural LGIs: Barguna Sadar Upazila and Barguna Sadar Union, located in the south-central part of the country closer to the centre area and good performer under UZGP and UPGP coverage respectively; and Kasba Upazila and Haripur Union, located in the east-central part of the country comparatively far from the centre area, and poor performance under UZGP and UPGP coverage respectively. In case of the ULGIs, we considered distance from the capital city, age of the institution and category according revenue collection performance as dimensions of variation. Distance/location of the municipalities and age of the institutions are taken as criteria because they are proxies for institutionalisation. Municipal categories Table 2 Bangladesh Institutional age Older Newer Bangladesh within-country case selection logic Rural 1 Upazila Parishad (sub-district) 1 Union Parishad 1 Upazila Parishad (sub-district) 1 Union Parishad Urban 1 Municipality 1 Municipality (A C) are based on the minimum annual revenues collected over the last three years. According to these criteria, we selected two urban LGIs: Moulavibazar municipality far from the capital city, long established and a Category A municipality; and Basail municipality closer to the capital city, relatively newly established and a Category B municipality. 2.3 Sector focus As mentioned earlier, the project also applies a comparative approach across a limited set of sectors and/or public functions to enable the disaggregated and inductive approach outlined above. While corruption may be widespread, and local governance important across many sectors, without exploring specific manifestations our approach posits that the resultant understandings of corruption and ways to fight it will risk being too generic to be actionable. We selected basic education as the sectoral focus for this project, with an emphasis on primary schooling. In addition, the project also looked at public financial management (PFM) as a core local government function. PFM is here understood as both the process and outcomes of management of public resources available to the local level. This process includes revenue management and planning, budget formulation and approval, budget execution, reporting and evaluation (Andrews et al., 2014). We chose these foci to enable common analysis across countries (both countries feature School Management Committees, for example), their impact on development outcomes for ordinary people, and the size of the public sector they represent. In Bangladesh, we also looked at selected social safety net services at the local level. 2.4 Data Due to the qualitative, small-n, and sensitive nature of the study, research design focused on triangulation (i.e. identifying potentially complementary sources of data). Four main categories of data were used, including desk reviews of existing country-focused literature, documentary and institutional sources, primary interviews, and focus group discussions. The interviews and focus groups used a semi-structured approach based on the detailed questions outlined in Annex 1, and analysis was via coding of responses to these research questions. To support comparable research approaches and tools, a checklist of likely data sources was agreed for both countries (Box 3). Details of data collection and fieldwork are available in the respective country studies. The next chapter details the state of theoretical and empirical understanding of decentralised governance and corruption, and argues that there are important theoretical perspectives in the literature on each concept 13

14 Figure 3 Bangladesh study sites Panchagarh Municipality Upazila Union Thakurgaon Nilfamari Lalmonirhat Dinajpur Rangpur Kurigram Basail Municipality Gaibandha Jaipurhat Sherpur Nawabganj Naogaon Bogra Jamalpur Mymensingh Netrokona Sunamganj Sylhet Rajshahi Nator Sirajganj Tangail Kishoreganj Habiganj Maulvibazar Maulvibazar Municipality Gazipur Pabna Narsingdi Kushtia Manikganj Brahmanbaria Meherpur Narayanganj Rajbari Dhaka Chuadanga Jhinaidaha Faridpur Munshiganj Magura Comilla Shariatpur Chandpur Jessore Madaripur Narail Gopalganj Lakshmipur Feni Barisal Noakhali Shatkhira Pirojpur Jhalokati Bagerhat Khulna Bhola Patuakhali Barguna Haripur Union Khagrachari Rangamati Chittagong Kasba Upazila Bandarban Cox's Bazar Barguna Sadar Union Barguna Sadar Upazila 14

15 that have not been adequately brought to bear on the links between them. Chapter 4 presents the study s empirical findings on local corruption and the dynamics of decentralised governance that may help explain them. It also explores the impact of local corruption in the sectors studied and its effects on the poor and marginalised. Chapter 5 reviews the effectiveness of anti-corruption efforts in mitigating these effects. The report concludes with implications for improving anticorruption efforts. Box 3 Data sources Documentary sources In-country literature/desk review, with specific focus on primary education and PFM Media scanning within selected local settings Primary institutional sources launch, budget, budget composition, establishment, meetings of council Interview sources Elected officials councillors, members of parliament Local government staff Sector staff and service providers project staff, contractors, delivery Citizens constituency based focus groups Activists and civil society organisations (CSOs) Local media representatives locality press club Political party representatives Other formal and informal decision-makers Expert groups academics, Transparency International, retired officials Local government ministry Education ministry 15

16 3 Understanding decentralisation and corruption A significant number of theoretical and empirical studies have addressed the relationship between decentralisation and corruption. However, these have yielded different interpretations of the relationship and evidence regarding decentralisation s effect on corruption. One reason for this diversity is that expectations and evidence on linkages between decentralisation and corruption are shaped by different theoretical perspectives adopted by analysts and policy-makers, and that there are gaps in these perspectives. Studies have also been focused on the impact of decentralisation reforms on corruption, but not on the converse role of corruption in shaping the logic and effectiveness of decentralised governance. This chapter reviews these different perspectives and identifies gaps in the way they have been used in relevant literature. 3.1 Empirical evidence on decentralisation, corruption and anti-corruption The empirical evidence from systematic studies of decentralisation and corruption is mixed, reinforcing the need to apply more nuanced theoretical and conceptual approaches. Several recent meta-analyses of anti-corruption evidence that capture decentralisation have confirmed that, while there is a large evidence base, findings on the links between decentralisation and corruption remain inconclusive (Rocha Menocal and Taxell, 2015; Johnson et al. 2012; Kolstad, Fritz and O Neil, 2008). Similarly, meta-reviews of decentralisation s impact on a range of development outcomes have found mixed evidence for its effects not only on corruption, but also on other variables of interest (Faguet and Pöschl, 2015; Local Development International, 2013; Mansuri and Rao, 2013), including poverty (Jütting, Corsi and Stockmayer, 2005), public service delivery (Kahkonen and Lanyi, 2001), and conflict management (Bird, 2003). These reviews capture large-scale quantitative and cross-country analyses (Shah, 2006; Fisman and Gatti, 2002; Fan, Lin and Treisman, 2009) that take broad approaches to conceptualising and operationalising both decentralisation and corruption. Typically they focus on fiscal shares of subnational expenditure or revenue and perception data on corruption, while a few deal with locallevel corruption (Albin-Lackey, 2007). As summarised by Fan, Lin and Treisman (2009), Huther and Shah (1998), De Mello and Barenstein (2001), Fisman and Gatti (2002) and Arikan (2004) all find that a larger subnational share of public expenditure was associated with lower perceived corruption. However, other studies report conditional effects, whereby larger subnational revenues are associated with lower perceived corruption only where parties are more established, or where there are fewer parties in government (Enikolopov and Zhuravskaya, 2007). Some studies focus on the effect of political arrangements: Goldsmith (1999), Treisman (2000), and Kunicová and Rose Ackerman (2005) all found that federal structure was associated with higher perceived corruption. (Fan, Lin and Treisman, 2009: 14) An additional problem with connecting the available research to concrete measures to fight corruption is the finding that perception indicators of corruption are not highly correlated with experiential indicators (Treisman, 2007b). Case studies and comparative work also emphasise that the interaction varies in different conditions. For example, a study of political decentralisation and corruption in Bangladesh, Côte d Ivoire, Ghana and India (Karnataka state) found that in the short run, decentralisation reduces grand theft but increases petty corruption, while in the long run, both may be reduced (Niazi, 2011). A few studies have touched on organisational and political cultures which might be influenced by decentralisation, with one example finding that: A move from command-and-control culture to service-oriented culture at a magnitude of one standard deviation may decrease corruption by over 17 percent. Civic institutions may also play an important role in anti-corruption efforts (16 percent). (Gurgur and Shah, 2005: 18) 16

17 Decentralisation and corruption are complex, and each has been approached through a range of theoretical perspectives. These perspectives are themselves evolving in relation to the elaboration of empirical evidence. Similarly, the empirical evidence for how decentralisation affects corruption depends on the theoretical approaches that are chosen to specify and measure decentralisation and corruption. These theoretical differences can help explain the seemingly contradictory and inconsistent evidence base on decentralisation and corruption. It is important, therefore, to understand the dominant theoretical perspectives on decentralisation and corruption and how they are changing. 3.2 Approaches to understanding decentralisation Decentralisation has well-established general definitions, though individual phrasing varies. For our purpose, as noted in Chapter 2: Decentralisation is generally understood as the assignment of public functions to subnational governments along with structures, systems, resources, and processes and that support implementing these functions to meet specific public sector goals. (Smoke, 2015a) However, behind this description there are many dimensions and forms of decentralisation. Generally these are divided into political, administrative and fiscal dimensions, and forms of decentralisation including devolution, deconcentration and delegation. These dimensions and forms combine to create a range of possible decentralisation scenarios, and it is important to disaggregate these to better understand their implications for development outcomes. For example, Bangladesh and Nigeria have very different decentralised governance forms. While Nigeria is structured around devolved relationships across three tiers of government (within a strongly devolved federal state relationship), Bangladesh features a generally de facto deconcentrated system, with strong ties to the centre, overlain with de jure devolution via some political decentralisation. There are three broad theoretical approaches to understanding decentralisation reforms (Table 3). Each builds on previous work, but should be considered three complementary lenses on decentralisation. Our fieldwork supports a combination of these approaches, and emphasises the importance of the third approach, one which focuses on the real-world implementation of Box 4 Dimensions and forms of decentralisation Administrative decentralisation is the transfer of managerial responsibility for functions (e.g. delivering a service) to: lower levels within existing departments (Administrative Deconcentration); levels of government (Administrative Devolution); or to non- or para-state entities (Administrative Delegation). Fiscal decentralisation is the provision of the resources for exercising decentralised powers and responsibilities to these levels. Political decentralisation is the shift of political decision-making power (e.g. legislative authority) to lower levels of the state, and increased accountability of these local levels to citizens through elections and/or other means (Political Devolution). It is sometimes referred to as democratic decentralisation. Source: European Commission (2007) Table 3 Three approaches to decentralisation Fiscal federalism Political economy Implementation and systems Key theories Key assumptions Implications for corruption research public choice principal agent rational agents benevolent principals understand incentives of subnational actors political economy new institutional economics incentives at all levels politics drives policy systems theories/multi-level approaches collective action institutional coherence and environment important inter-government relationships important understand incentives at all levels understand corruption systems understand social/collective factors understand corruption effects on decentralisation 17

18 decentralisation but which is relatively neglected in the literature connecting decentralisation and corruption Fiscal federalism The first approach derives from theories of public economics and public choice. It focuses on how decentralisation of the public sector affects its three economic functions of allocation, distribution and stabilization (Oates, 1989: 554). 1 A lot of effort in this approach is devoted to the assignment problem that is, what distribution of functions across levels of government is likely to create the best outcomes? Stabilisation functions, such as control of the money supply, are best held centrally. However, allocation and distribution are better served by different degrees of fiscal federalism. This proposition was captured in Wallace Oates decentralisation theorem (1972), which posits that different spatial characteristics of public goods mean that some will be better provided by local jurisdictions. Some goods, such as defence or those with high economies of scale, are best produced nationally, while other goods benefit only specific areas (have few externalities) or are subject to preferences that differ in different places. In these cases, allocation or efficiency can be improved through decentralisation, or competition among jurisdictions can improve efficiency (Tiebout, 1956). Similar principles are applied to the assignment of revenues for example, taxing mobile factors locally does not make sense, so property tax is an ideal local revenue source (Tanzi, 1995: ). This approach leads to a prescriptive and sequenced model for decentralisation that starts with the functions to be carried out by each level of government, the best assignment of revenues to these levels, then determines how transfers and borrowing can best fill the gaps (Steffensen, 2010: 5). This approach has been called First Generation Fiscal Federalism (Weingast 2014). It frames decentralisation as a way to improve public sector performance within principal agent relationships. As such, it implies Box 5 Four pillars of fiscal decentralisation expenditure assignment revenue assignment transfers borrowing Box 6 Principal agent problems A principal agent problem exists when one party to a relationship (the principal) requires a service of another party (the agent) but the principal lacks the necessary information to monitor the agent s performance in an effective way. The information asymmetry that arises from the agent having more or better information than the principal creates a power imbalance between the two and makes it difficult for the principal to ensure the agent s compliance (Booth, 2012). that principal agent analysis is the primary tool to understand the relationship between decentralisation and outcomes like corruption, and it aligns closely with the principal agent approach to corruption described later in this chapter Political economy A second approach to decentralisation emphasises political economy, and loosens the assumption of a benevolent Principal: Views public officials as conventional utility maximizing entities with their own objective functions, who operate within a constellation of incentives and constraints that depends on existing political and fiscal institutions. (Oates 2008) This approach focuses on the interests and motivations of those initiating or driving decentralisation processes, as well as those affected by them. Decentralisation almost never begins through an ordered policy process (Smoke, 2015b: 252). Rather, it is a response to political conditions, through which its progress is also shaped. A key feature of the late 20th and early 21st century has been a widespread tendency to adopt decentralisation reforms. 2 However, this trend masks great diversity in the causes, drivers and pathways of such reforms. For example, rapid decentralisation of a broad range of functions in Indonesia and subsequent strengths and weaknesses cannot be isolated from the context of separatism, fiscal crisis, and democratic transition that surrounded this big bang reform. Similarly, Nigeria s dispensation cannot be understood without reference to historical concerns over regionalism, diversity and national unity, and the need to manage rents from resource revenues. And of course, 1 These functions are also sometimes referred to in public financial management literature as macroeconomic stabilisation or fiscal discipline, technical efficiency, and allocative efficiency (World Bank, 1998: 10; Allen and Tommasi, 2001: 19 20). 2 Various analysts have suggested decentralisation reforms were on the agenda in between 80%and virtually all countries during this period (World Bank, 1999: ; Faguet, 2014). 18

19 decentralisation is shaped by historical starting points colonial legacies, legal origins and the like. Analysts have classified some of the political factors that shape the initiation and course of decentralisation processes. Some causes identified in the literature include: 3 prevention or mitigation of subnational conflict, particularly ethnic conflict alleviation of fiscal pressure or crisis at national level consolidation of central government control or legitimacy improvement of service delivery stimulation of economic growth pressure and influence from donors. Different causes of decentralisation will emphasise different dimensions of decentralisation, producing variants in the type of decentralisation that results. This second approach focuses attention on the diversity of decentralisation, helping to distinguish the variants of decentralisation and the different effects these might have. Some of these effects appear in the empirical findings of this study. In Nigeria the strong emphasis on political and democratic decentralisation through devolution has created supposedly strong local governments but without adequate fiscal or administrative autonomy to act on their assigned role. In Bangladesh, administrative decentralisation has also run ahead of fiscal dimensions, and incomplete political decentralisation means accountability relationships remain confused. For example, depending on the political economy driving decentralisation, one or another of the political, administrative and fiscal dimensions of decentralisation may dominate. What this perspective implies or in some cases explicitly concludes is that if the dimensions of a decentralisation process become too far out of balance, it is unlikely to function well (Boex, n.d.). Large revenueled fiscal decentralisations can result in the transfer of resources in the absence of adequate capacity to plan, budget and monitor. Politically-led decentralisation can lead to unfunded mandates : responsibilities devolved to local levels without adequate funding. Alternatively, administrative decentralisation without clear accountability may produce predatory behaviour by empowered local officials. This second approach focuses attention on the diversity of decentralisation, helping to distinguish the variants of decentralisation and the different effects these might have. Some of these effects appear in the empirical findings of this study. In Nigeria the strong emphasis on political and democratic decentralisation through devolution has created supposedly strong local governments but without adequate fiscal or administrative autonomy to act on their assigned role. In Bangladesh, administrative decentralisation has also run ahead of fiscal dimensions, and incomplete political decentralisation means accountability relationships remain confused Implementation and systems approach The third approach to decentralisation emphasises the process of reform and brings to the fore gaps between policies and implementation. While there is an optimal sequencing for decentralization countries in general, do not honour this sequence often for the political economy reasons highlighted in the second approach (Bahl and Martinez-Vazquez, 2006: 34). Changes to central local relations empower new actors, create new interests and interest groups, and engender resistance. These challenges are linked to the political process by which decentralisation is driven, but also stem from a range of neglected issues including contextual factors, coordination problems and bureaucratic dynamics among the actors involved, and variations in local and national capacity (Smoke, 2013; 2015b: 253). In this view, a key to understanding the variable outcomes of decentralisation lies in this implementation challenge (Smoke, 2010: 192). This approach emphasises gaps between national and subnational governments in information, finances, capacity, policy and political orientation and other areas (Charbit, 2011b: 15 16). The quality of the relationships and coordination between levels of government is as important as whether a country should decentralize or not (Charbit, 2011b; Smoke, 2015c; Mansuri and Rao, 2013). Capacity, coherent policies, institutions and Table 4 Causes, forms and manifestations of different decentralisation processes Driver Dominant dimension Manifestation Pressure to reduce central government budgets Fiscal Assignment of expenditure functions to lower levels (with or without funding) Prevention of subnational conflict Political Creation of subnational government representation Consolidation of central control Administrative Creation of deconcentrated administrative units Efficiency/services Delegation New public management/donor-driven local capacity development 3 For discussions of the drivers of decentralisation see Eaton, Kaiser and Smoke (2011), and Smoke (2015a). 19